JP2021517684A - Managing cache behavior using track locks and stride group locks - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a computer program product, a system, and a method for managing cache operation by using a track lock and a stride group lock. A group of tracks from a storage device is stored in a cache. For writing to a track in a group in the cache, an exclusive track lock is granted to the track in the group in the cache, and the exclusive track lock is for writing to a different track in the cache. Can be held at the same time. An exclusive group lock is granted to the group of tracks in the cache, and the tracks in the group are destaged from the cache to the storage device. The exclusive group lock is released in response to completing the destage of the trucks in the group in the cache to the storage device. [Selection diagram] Fig. 3

Description

The present invention relates to computer program products, systems, and methods for managing cache operation using track locks and stride group locks.

The storage controller maintains one or more device adapters for interfacing with an array of storage devices, such as an array of independent disk redundant arrays (RAIDs) of the storage device. The device adapter manages the requests to the attached storage device and implements the RAID controller and functions for the storage controller. In Nonvolatile Memory Express (NVMe), NVMe controllers access read and write requests to storage devices in the host system's submission queue to access solid state drives (SSDs:). It may include a RAID function that strips data between storage devices such as Solid State Drive).

Improved for reading and writing data across an array of storage devices such as RAID arrays in an environment where multiple processor cores read and write to the tracks of the stride group in the cache stored in the storage device. Technology is needed in this field.

The first embodiment provides computer program products, systems, and methods for managing cache operations using track locks and stride group locks. A group of tracks from the storage device is cached. An exclusive track lock is granted to a track in the group in the cache for writing to a track in the group in the cache, and an exclusive track lock is granted for writing to a different track in the cache. Can be held at the same time. An exclusive group lock is granted to the group of tracks in the cache, and the tracks in the group are destaged from the cache to the storage device. The exclusive group lock is released in response to completing the destage of the trucks in the group in the cache to the storage device.

The first embodiment provides improvements to computer technology that allow multiple exclusive track locks for writing to be held simultaneously, allowing writing to different tracks in the cache. In addition, an exclusive group lock is granted to the group of trucks due to the destage of the group of trucks. While an exclusive group lock is held for the destage process, different processes attempting to access the destaged track are blocked. When the exclusive group lock for destage is released, it becomes possible to grant the exclusive track lock for simultaneous writing to the tracks in the group. Thus, the described embodiment manages and enables simultaneous writing to different tracks within a group of tracks, such as a stride group, and locks all tracks within the group for destage operation. Provide improvements to provide.

The second embodiment further optionally receives a write to the target track in the group, determines whether an exclusive group lock to the group is held, and an exclusive group lock. In response to determining that is not held, granting an exclusive track lock to the target track, and in response to determining that an exclusive group lock is held, for the target track It may include retrying access to the exclusive track lock.

In the second embodiment, writing to the target track is permitted only when the exclusive group lock for the destage operation is not held, and the writing is a track involved in the ongoing destage operation. Do not write to. This provides improvements for coordinating simultaneous writing to different tracks within a group of tracks and destage behavior to a group of tracks for destage all tracks within the group.

The third embodiment optionally further grants a shared group lock for writing to the tracks in the cache and in response to the completion of writing to the tracks in the cache. Releasing the shared group lock for writing and determining if any shared group lock is held in response to initiating the destage of a group of tracks. An exclusive group lock may include determining that it is granted only in response to determining that there is no shared group lock held for the destaged group.

In a third embodiment, the shared group lock is granted for writes that simultaneously write to tracks in the group to which the shared group lock is applied. To further adjust the destage operation while simultaneous writes are occurring, the shared group lock held by the write prevents the destage operation from starting until all writes are complete, which is destage. This is indicated by the fact that there are no shared group locks held for the group of tracks that you want to play.

In the fourth embodiment, the destage of the group of tracks further identifies the modified data in the tracks in the group in the cache, and the altered data and the old parity data of the group. Read from at least one of the storage devices and stage it in the cache by generating a read request indicating the data about and adding it to the submission queue, and the modified data staged in the cache. And using the read data about the old parity data to calculate the new parity data and at least one of the new parity data and the changed data in the cache on the storage device. The exclusive group lock is released in response to the completion of writing the new parity data and the changed data to at least one of the storage devices. It can include writing and including.

A fourth embodiment generates a read request to read the modified and old parity data from the storage device, stages it in the cache, and then uses the staged data to create a new parity. Computing the data provides an improvement in the calculation of the parity data of the tracks in the group. The destage operation holds an exclusive group lock while computing the new parity data, which is not released until the new parity data is calculated and written to the storage device. This improves coordination between destaging behavior and writing to the track by preventing writes until all modified and new parity data on the track is destaged.

In a fifth embodiment, optionally, writing the new parity data and the changed data generates a write request indicating the changed data and the data about the new parity data in the cache, and the sub. By adding to the mission queue, it may include transferring from the cache to at least one of the storage devices.

In a fifth embodiment, a write request for writing changed data and new parity data in the cache is added to the submission queue and transferred from the cache to the storage device. This write request management deploys a controller that accesses read and write requests from the submission queue, such as in the case of NVMe controllers, where the controller writes only changed and new parity data to storage. It is useful in systems that allow the read and write of modified and parity data for partial group destage operation to be completed. In this way, the bandwidth between the controller and the system is optimized by transferring only the changed data and the new parity data through the submission queue.

A sixth embodiment optionally generates a read request to identify a modified segment of a track in a group and cache segments in the cache, identified modified segments and old. Assigning to a segment of parity data and generating a list showing the segment of modified and parity data and the destination location in the cache of the allocated cache segment, the list is storage. It may include generating, provided with a read request, for use in identifying data to be read from at least one of the devices and written to a destination location in the cache.

A sixth embodiment is a segment of modified and parity data by generating a list showing the segments of modified and old parity data provided with the read request in the submission queue. Improves the efficiency of staging operations to the cache. The controller can access the submission queue to process read requests and lists, stage modified and old parity data segments into the cache, and use them to calculate new parity data. A controller, such as an NVMe controller, accesses a read request from the submission queue to stage the segments needed to calculate the new parity data. The described embodiment optimizes the staging behavior by staging only the segments of the data that have changed, not the segments that have not changed, in partial stride destage.

A seventh embodiment provides computer program products, systems, and methods for managing input / output requests to tracks in caches stored in multiple storage devices, which is a plurality of processes. This is done by the core executing program code in computer-readable storage to grant a shared stride lock to writes to the tracks in the stride group of tracks, which is a stride lock. Granted for writes that are written to different tracks in the group at the same time. In response to a destage request to destage a track's stride group, a determination is made as to whether there is an unreleased shared stride lock for writing. An exclusive stride lock is granted for destage of the track stride group from the cache to the storage device in response to determining that there are no unreleased shared stride locks for writing.

A seventh embodiment is improved caching by granting a shared stride lock to the processing cores that write to different tracks in the stride group at the same time when multiple processing cores write to tracks in the stride group. Providing technology. The destage request cannot proceed until an exclusive stride group lock has been granted, which cannot be granted until there are no unreleased shared stride locks for writing. Thus, the described embodiments include writing by multiple cores that can write to tracks in the stride group at the same time and destage operation of the stride that requires exclusive access to all tracks in the stride group. Provides improved technology for adjusting.

It is a figure which shows one Embodiment of a storage controller. It is a figure which shows one Embodiment of a truck lock. It is a figure which shows one Embodiment of the operation for processing the read request of the track in a stride group. It is a figure which shows one Embodiment of the operation for processing the write request to the track in a stride group. It is a figure which shows one embodiment of the operation for performing a partial stride destage of a track in a stride group. It is a figure which shows one embodiment of the operation for performing the complete stride destage of a track in a stride group. It is a figure which shows the computing environment in which the component of FIG. 1 can be implemented.

The described embodiments provide cache lock management techniques for improving read and write management by multiple processor cores for stride group tracks striped between storage devices. In certain embodiments, the storage controller does not have to include a device adapter to handle RAID operations, instead the submission queue is accessed by the controller to read and write to the storage device. Reads and writes may be placed in. The described embodiments include management of reads and writes to the cache by multiple processor cores and within a group stored in an array of storage devices, such as stride groups striped between storage devices in a RAID array. Provides a track destage and. The described embodiments provide an exclusive track lock that allows an independent core or process to write to different tracks in a stride group at the same time, and is exclusive for managing the destage of tracks in the stride. And provide shared group locks. The destage operation may only be allowed to proceed if an exclusive group lock is received, the exclusive group lock has other writers holding the lock to the tracks in the stride group. Granted if not, and if granted, prevent writing to the stride group during destage operation.

The described embodiments provide improvements in cache management techniques in an environment where a separate controller manages reads and writes between the cache and the storage device. The described embodiment makes it possible to adjust the locked write to write to a track in the stride group so that the write does not proceed during the destage of the stride group being written across the storage device. ..

FIG. 1 shows a plurality of cores 102 ₁ , 102 ₂ . .. .. .. It _{illustrates an embodiment of a storage controller 100 or other type of system that includes 102 n} and memory 104, where memory 104 caches staged tracks from storage device 108 in storage array 110. A cache 106 is included so that it can be used in a read request. The storage controller 100 communicates with the subsystem 112 via the fabric 114, and read and write requests to the storage device 108 are managed by the subsystem 112. The memory 104 has cores 102 ₁ and 102 ₂ . .. .. .. Each of the 102 _n contains one queue 116, and each core queue 116 includes one or more submission queues (SQs) 118 to which read and write requests are added. Includes completion queue 120 (CQ: completion queue). Queues 118 and 120 have a head pointer that points to one end of the queue from which the request is accessed when the queue is processed, and a tail pointer that points to one end of the queue to which the request is added. May have a queue. Core 102 _i executes the I / O manager 122 so that core 102 _i sub-stages the track from storage device 108 to cache 106 and destages the track from cache 106 to storage device 108. Add to mission queue 118.

If the I / O manager 122 writes to the submission queue doorbell 126 in the controller 124, the controller 124 in the subsystem 112 is a storage device added to the submission queue 118 by the I / O manager 122. The read request and the write request for 108 are taken out. In response to completing the fetched read or write request to the storage device 108, the controller 124 returns the completion of the read or write request to the completion queue 120 and initiates the read or write request. 102 ₁ , 102 ₂ . .. .. .. Notify 102 _n of completion. _{The core 102 i} running the I / O manager 122 may adjust the completion by writing to the completion queue doorbell 128 to indicate that it has finished processing the completion written to the completion queue 120.

In a non-volatile memory express (NVMe) embodiment, the controller 124 may include an NVMe controller, and the I / O manager 122 may include an NVMe driver for interfacing with the NVMe controller 124 and implementing the queue 116. NVMe controller 124, one for each submission queue used by the core 102 _i, includes a submission queue doorbell 126 also include a completion queue doorbell 128. _{The core 102 i,} which adds the request to its own submission queue 118, writes to the submission queue doorbell 126 of controller 124 and is associated with the submission queue doorbell 126 written to controller 124. Access read or write requests added to mission queue 118. Controller 124 may execute the fetched commands in any order. Core 102 ₁ , 102 ₂ . .. .. .. When 102 _n processes the completion of the read / write request indicated in the completion queue 120, cores 102 ₁ , 102 ₂ . .. .. .. 102 _n writes to the completion queue doorbell 128 of the completion queue 120. The controller 124 may support a plurality of namespaces 130 including each part of the non-volatile memory space mounted on the storage device 108. The storage controller 100 may direct read and write requests to namespace 130, which maps to the non-volatile memory space within the storage device 108. The storage device 108 will further include hardware and software for interfacing with the NVMe controller 124.

In the embodiment of FIG. 1, NVMe may be implemented via the fabric, and the storage controller 100 and the subsystem 112 communicate via the fabric 114. In such an embodiment, the storage controller 100 and the NVMe subsystem 112 are, for example, remote direct memory access (RDMA), RDMA over-converged Ethernet (R) (RoCE: RDMA over). It includes transport layers 132, 134 that enable network communication using Converged Ethernet), Internet Wide Area RDMA Protocol (iWARP: Internet Wide-area RDMA Protocol), InfiniBand, and Fiber Channel, respectively. In an alternative embodiment, the subsystem 112, such as the NVME subsystem 112, may be implemented within a Peripheral Component Interconnect Express (PCIe) card within the storage controller 100, the storage controller 100 and the NVMe subsystem. The 112 communicates via the PCIe interface.

Fabric 114 includes a local area network (LAN), a storage area network (SAN), a wide area network (WAN: Wide Area Network), a peer-to-peer network, and a wireless network. , The Internet, etc. may be provided with one or more networks.

_{The storage controller 100 has cores 102 1} , 102 _2. To stage and destage tracks from storage device 108 to cache 106. .. .. .. Holds the input / output (I / O: Output / Output) manager code 122 executed by 102 _n. I / O manager code 122 also implements an independent disk redundant array (RAID) algorithm or other type of storage array management algorithm to strip the tracks in the stride group to storage device 108 and stride. The parity of the tracks in the group can be calculated and striped with the tracks between the storage devices 108. The I / O manager 122 maintains a stride group lock 200 for a stride group configured within the storage device 108 to provide different cores 102 ₁ , 102 ₂ . .. .. .. By managing the reads, writes, staging, and destages of tracks within the stride group being performed by 102 _{n, conflicts are avoided and cache and destage consistency is maintained.}

Core 102 ₁ , 102 ₂ . .. .. .. The 102 _n may include a plurality of cores on the same central processing unit (CPU) / integrated circuit board, or may include separate processing units. Each core 102 ₁ , 102 ₂ . .. .. .. The 102 _n executes the code of the I / O manager 122 to send a read / write to the storage device 108 held in the cache 106 to the track composed of the stride, so that the track can be tracked from the storage device 108. Can be staged and the stride group of tracks destaged from the cache 106 to strip between storage devices.

The I / O manager 122 may hold a track index that provides the index of the track in the cache 106 to the cache management block in the management block directory. The management block directory contains cache management blocks, and there is one cache management block for each track in cache 106, which holds metadata about that track in the cache, for example, that track. Provides stride group information about stride groups. The track index associates a track with a cache management block to provide information about the track in cache 106.

The storage controller 100 is a storage system manufactured by International Business Machines (IBM (registered trademark of IBM): International Business Machines Corporation) DS8000 (registered trademark of IBM) and DS8880 storage system, or other vendors. It may have a storage system, such as a controller and a storage system. (IBM (registered trademark of IBM) and DS8000 (registered trademark of IBM) are trademarks of International Business Machines, Inc. worldwide.)

The storage device 108 in the storage array 110 may comprise a different type or class of storage device, eg, a solid state consisting of a magnetic hard disk drive, magnetic tape storage, and solid state electronics. Storage device (SSD), EEPROM (electrically erasable programmable read-only memory: Electrically Erasable Programmable Read-Only Memory), flash memory, flash disk, random access memory (RAM: Random Access Memory) drive , Storage class memory (SCM: studio-class memory), phase change memory (PCM: Phase Change Memory), resistance random access memory (RRAM: receive random access memory), spin injection memory (STM-RAM: spin) Transfer torque memory), conductive bridge RAM (CBRAM), magnetic hard disk drives, optical disks, tapes and the like. Volumes in storage can also include simple disk bundles (JBOD: Just a Bunch of Disks), direct access storage devices (DASD), independent disk redundant array (RAID) arrays, virtualization devices, and more. , Can consist of an array of devices. In addition, the storage device 108 in the storage array 110 has a higher data transfer rate than different types of storage devices from different vendors and different types of storage devices, such as second type storage devices such as SSDs. It may be equipped with a first type of storage device such as a low hard disk drive.

In an NVMe implementation, the storage device 108 will include NVMe components for interfacing with the NVMe subsystem 112, such as NVMe SSD devices.

The memory 104 may include suitable system memory known in the art, which includes volatile and non-volatile memory devices such as Dynamic Random Access Memory (DRAM). ), Magnetoresistive Memory (PCM), Magnetoresistive Random Access Memory (MRAM), Spin Transfer Torque (STT) -MRAM, SRAM Storage Device, DRAM, Intensifier Random Non-volatile direct in-line memory module with access memory (FeTRAM), nanowire-based non-volatile memory, and byte-addressable write-in-place memory. (DIMM: Non-Volatile Direct In-Line Memory Mode) and the like are included.

Figure 2 illustrates an embodiment of a stride group lock 200 _i instance to one of the configured stride group in storage device 108, which is, stride group identifier (ID for identifying the stride groups : Identifier) 202 and track locks for each track in stride group 202 204 ₁ , 204 ₂ . .. .. _{The track lock 204 i} for each track in the stride group 202, which is 204 _n , gives the track an exclusive track lock 206 indicating whether or not the track has an exclusive lock. _{Track locks 204 1} , 204 _2. Includes a shared track lock counter 208 that indicates the number of shared track locks that have been made. .. .. 204 _n , an exclusive stride lock 210 indicating that an exclusive lock has been granted to the stride group 202, a shared stride lock counter 212 indicating the number of shared stride locks granted, and including. Stride Group lock 200 _i information may further indicate the tracks stride in the group lock 204 _i is provided.

_{Figure 3 shows cores 102 1} , 102 ₂ for processing read requests for tracks in the stride group. .. .. .. An embodiment of an operation performed by the I / O manager 122 performed by 102 _{n is shown. Processor / Core 102 1} , 102 ₂ (at block 300). .. .. .. Upon receiving a read of a track in the stride group from 102 _{n, if the} requested track (in block 302) is in cache 106, the I / O manager 122 will use the track or stride 210 containing the track (in block 304). It is determined whether or not the exclusive track lock 206 is held with respect to. If held, the track is being written or destaged, and a read request (at block 306) is queued or retried to acquire a shared lock on the track. If exclusive track lock 206 or exclusive stride lock 210 is not held (at block 304), the track's shared track lock counter 208 is incremented (at block 308) to request cores 102 ₁ , 102. ₂ . .. .. .. A shared track lock is granted to _{102 n.} The shared track lock counter 208 is decremented when the read is complete (at block 310).

If the requested track (in block 302) is not in cache 106, the I / O manager 122 allocates a segment to the requested track in cache 106 (in block 312) and the requested track (in block 314). Is granted an exclusive track lock 206 and the shared stride lock counter 212 is incremented to grant a shared stride lock to the stride containing the staging track. A read request to read the requested track segment (at block 316) is queued to submission queue 118. In NVMe implementation, NVMe driver storage controller 100 writes the submission queue doorbell 126 through fabric 114 or bus interface, to process the submission queue 118 of the core 102 _i that initiated the read request The controller 124 may be signaled that there is an I / O request to be made. Controller 124 may write a segment of the track (at block 318) to the segment assigned to the track in the cache 106. The controller 124 also writes the completed core 102 _i of the completion queue 120 that initiated the request, signaling that the reading has been completed. The I / O manager 122 then releases the track lock 206 to the requested track (at block 320), decrements the shared stride lock counter 212, and shares it to the stride containing the staging track. By releasing the stride lock, the lock held on the staging track can be released. After releasing the locks held for the staging process, control proceeds to block 308, may allow to read the tracks from the cache 106 to start the core 102 _i.

In the embodiment of FIG. 3, the exclusive track lock 206 and the exclusive track lock 206 and before allowing the track and the stride group including the track to be staged in the cache for a read request to be written or destaged to that track. The shared stride lock is retained. The described embodiment states that track reads occur while different cores can write to different tracks in a stride group at the same time, or while the tracks are in a destaged stride group. Enables and provides improvements to computer technology. Moreover, in certain embodiments, destage to the stride group may be allowed to proceed while there is a pending read on the track within the stride group. The described embodiments provide locks, data structures, and functions for coordinating the behavior of tracks in the cache with the staging of the tracks into the cache.

_{Figure 4 shows cores 102 1} , 102 ₂ for processing write requests to tracks in the stride group. .. .. .. An embodiment of an operation performed by the I / O manager 122 performed by 102 _{n is shown. Processor / Core 102 1} , 102 ₂ (at block 400). .. .. .. Upon receiving a write from 102 _n to a track in the stride group, the I / O manager 122 asks whether an exclusive 206 or shared 208 track lock is held on the track (at block 402), or It determines whether there is an exclusive stride lock 210 held against stride group 200 _i including the tracks to be written. If held, the track is being read, written, or destaged, and a write request is queued or retried (at block 404) to acquire an exclusive lock 206 on the track. If the exclusive 206 / shared 208 track lock or exclusive stride lock 210 is not held (at block 402), then the exclusive track lock 206 (at block 406) is the request core 102 ₁ , 102 ₂ . .. .. .. Granted to 102 _n , the shared stride lock counter 212 is incremented (at block 408) to grant shared stride lock to the stride group containing the track to write. A write is then performed to the allocated track in cache 106 (at block 410). When the write is complete, the exclusive track lock 206 is released (at block 412), the shared stride lock counter 212 is decremented, and the shared stride lock is released.

In the embodiment of FIG. 4, the track is written while the track is being written to the cache 106 by holding an exclusive track lock and a shared stride lock for the track to be written and the stride group including the track. Is not read, or the stride group containing the track is not destaged. The described embodiments allow writes to a track to occur simultaneously while different cores are writing to tracks in a stride group at the same time, and the track or device subject to pending writes. By not initiating writing to the tracks within the stride group being staged, it is possible to coordinate the destage of the stride group with the writing behavior, providing improvements to computer technology. In addition, the described lock embodiments are for tracks within the stride group, as indicated by the shared stride lock being held, which is indicated by the shared stride lock counter 212 greater than zero. Prevents destage from progressing while there are pending writes.

_{FIG. 5, cores 102 1} , 102 ₂ to destage a partial stride group having, but not all, tracks in the stride group in the cache 106. .. .. .. An embodiment of an operation performed by the I / O manager 122 performed by 102 _{n is shown.} The I / O manager 122 (at block 500) is known in the art as a selection algorithm for selecting stride groups to be destaged, such as Wise Order Writing (WOW). Destage movements can be initiated using other suitable stride group selection techniques and the like. Exclusive stride lock 210 can now be granted if a shared stride lock is held for a stride group, as indicated by the shared stride lock counter 212 (in block 502) being greater than zero. The destage request is queued or retried (at block 504) until. If the stride group is not held against the destaged shared stride lock (in block 502), i.e. if the shared stride lock counter 212 is zero, then the I / O manager 122 (in block 506) Grants an exclusive stride lock 210 to the destaged stride group. The modified track of the stride group in cache 106 is identified (at block 508). A cache segment is assigned to the identified modified track and old parity data in cache 106 (at block 510) to staging the modified version and old parity data of the modified track from storage device 108. Will be available to do. The I / O manager (at block 512) may generate a read list of modified track segments in storage device 108 and segments with old parity data, staging to the allocated cache segment. .. In an NVMe implementation, the read list may include a distributed collection list (SGL) or a physical region space (PRP) list showing the segments read from the storage device 108 into the cache 106. The I / O manager 122 (at block 514) has a segment of modified track and old parity data to read and a destination position in the cache 106 of the allocated cache segment to store the read segment. A read request (or a plurality of read requests) having a read list that identifies is added to the submission queue 118 of the core 102 performing the destage operation. The I / O manager 122 writes to the doorbell register (at block 516), the submission queue doorbell 126, and signals the controller 124 to read the submission queue 118 to which the read request has been added.

Completion queue for _{core 102 i} where controller 124 is destage (at block 518) after controller 124 writes a segment in the read list from storage device 108 to an allocated cache segment in cache 106. Adding completion to 120 causes the I / O manager 122 (at block 520) to run during read-modify-write using the modified track, the old data of the modified track, and the old parity. A new parity is calculated by XOR calculation for the data, and the parity is updated. After updating the parity, the I / O manager 122 (at block 524) has a write list showing the cache 106 segments with the modified track of the destaging stride group and the calculated new parity, eg, In the case of NVMe implementation, a distributed collection list or PRP list is generated. The I / O manager 122 (at block 526) has a write request (or a plurality) with a write list that identifies a cache segment of modified tracks and new parity data to write to namespace 130 in storage device 108. Write request) is added to submission queue 118. The I / O manager 122 writes to the doorbell register (at block 528), the submission queue doorbell 126, and signals the controller 124 to read the submission queue 118 to which the write request has been added.

After controller 124 writes a segment in cache 106 shown in the write list to the destination location of storage device 108, the completion queue 120 of _{core 102 i where controller 124 is destaged (at block 530).} Adding completion to the I / O manager 122 releases the exclusive stride lock 210 to the destaged stride (at block 532).

In the embodiment of FIG. 5, an exclusive stride lock is held for the destaged stride group to prevent the track from being read or written while the track is being destaged, but during the destage operation. , Read hold may continue. The described embodiment prevents reading and writing from starting during destage operation and uses an unmodified version of the modified track used to update the parity when all tracks in the destaged stride group are not in the cache. By staging segments, it provides improvements to computer technology for destaging partial stride groups. On the other hand, there are pending writes to the tracks in the stride group, as indicated by the shared stride lock being held, which is indicated by the shared stride lock counter 212 being greater than zero.

_{Figure 6 shows cores 102 1} , 102 ₂ to destage the full stride group when all tracks in the stride group, including the modified track, are in cache 106. .. .. .. An embodiment of an operation performed by the I / O manager 122 performed by 102 _{n is shown.} The I / O Manager 122 (at block 600) has a selection algorithm for selecting the stride group to be destaged, such as Wise Order Writing (WOW) or other suitable stride known in the art. -The destage operation can be started by using a group selection technique or the like. Exclusive stride lock 210 can now be granted if a shared stride lock is held for the stride group, as indicated by the shared stride lock counter 212 (in block 602) being greater than zero. The destage request is queued or retried (at block 604) until. If the stride group is not held against the destaged shared stride lock (in block 602), i.e. if the shared stride lock counter 212 is zero, then the I / O manager 122 (in block 606) Grants an exclusive stride lock 210 to the destaged stride group. The I / O manager 122 then calculates the parity from all tracks of the stride in the cache 106 (at block 608) and updates the parity with the new parity.

The I / O manager 122 produces a write list showing the segment of the stride modified track (at block 610) and the calculated new parity. The I / O manager 122 submits a write request with a write list that identifies the changed track and the cache segment of new parity data to write to the namespace in storage device 108 (at block 612). Add to queue 118. The I / O manager 122 writes to the submission queue doorbell 126 (at block 614) and signals the controller 124 to read the submission queue 118 to which the write request has been added.

The segments in the cache 106 by the controller 124 is shown to the write list after writing to the destination location of the storage device 108, the completion queue 120 of the core 102 _i the controller 124 is running (block 616) destage Adding completion releases the exclusive stride lock 210 to the destaged stride (at block 618).

In the embodiments of FIGS. 5 and 6, holding an exclusive stride lock on the destaged stride group prevents the track from being read or written while it is being destaged, but destaged. The read hold may continue during operation. The described embodiment prevents reading and writing from starting during destage operation and uses an unmodified version of the modified track used to update the parity when all tracks in the destaged stride group are not in the cache. By staging segments, it provides improvements to computer technology for destaging partial or complete stride groups.

The present invention may be implemented as a system, method, or computer program product or a combination thereof. A computer program product may include a computer-readable storage medium (or a plurality of media) on which computer-readable program instructions for causing a processor to perform aspects of the invention.

The computer-readable storage medium can be a tangible device that can hold and store instructions for use by the instruction execution device. The computer-readable storage medium can be, for example, but not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination described above. A non-exhaustive list of more specific examples of computer-readable storage media includes portable computer disksets, hard disks, random access memory (RAM), read-only memory (ROM), Erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM: static random access memory), portable compact disk read-only memory (CD-ROM: portable compact disk read-only) Mechanically coded such as memory), digital versatile disc (DVD: static random disk), memory stick (R), floppy disk (R), punch card with instructions or raised groove structure. Devices, and any suitable combination described above. Computer-readable storage media, as used herein, are, for example, radio waves or other freely propagating electromagnetic waves, waveguides or other electromagnetic waves propagating through transmission media (eg, optical pulses through fiber optic cables). ), Or an electrical signal transmitted by wire, should not be construed as a transient signal itself.

The computer-readable program instructions described herein are from computer-readable storage media to their respective computing / processing devices, or, for example, the Internet, local area networks, wide area networks, or wireless networks, or theirs. It can be downloaded to an external computer or external storage device via a network such as a combination. The network may include copper transmission cables, optical transmission fibers, wireless transmissions, routers, firewalls, switches, gateway computers, or edge servers or combinations thereof. The network adapter card or network interface of each computing / processing device receives computer-readable program instructions from the network, transfers the computer-readable program instructions, and stores them in the computer-readable storage medium in each computing / processing device. do.

Computer-readable program instructions for performing the operations of the present invention include assembler instructions, instruction set-architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, Alternatively, in any combination of one or more programming languages, including object-oriented programming languages such as Java (R), Smalltalk, C ++, and traditional procedural programming languages such as the "C" programming language or similar programming languages. It can be written source code or object code. Computer-readable program instructions are entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on the remote computer, or completely. It can run on a remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or wide area network (WAN), or (eg, the Internet). • Can be connected to an external computer (via the Internet using a service provider). In some embodiments, electronic circuits including, for example, programmable logic circuits, field programmable gate arrays (FPGAs), or programmable logic arrays (PLAs) are described in the present invention. A computer-readable program instruction can be executed by utilizing the state information of a computer-readable program instruction for personalizing an electronic circuit in order to execute the above-described embodiment.

Aspects of the invention are described herein with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the invention. It will be appreciated that each block of the flow chart and / or block diagram, and the combination of blocks in the flow chart and / or block diagram, can be implemented by computer-readable program instructions.

These computer-readable program instructions are provided to a general purpose computer, a dedicated computer, or the processor of another programmable data processor, and the instructions executed through the processor of the computer or other programmable data processor are shown in the flowchart. Alternatively, a machine may be created that produces means for performing a specified function / action in one or more blocks of a block diagram or both. Also, these computer-readable program instructions include instructions in which the computer-readable storage medium in which the instructions are stored performs the mode of function / action specified in one or more blocks of the flowchart and / or block diagram. It may be stored on a computer-readable storage medium that can be directed to a computer, a programmable data processor, or other device or a combination thereof to function in a particular way, including a product.

A computer can also be loaded with computer-readable program instructions into a computer, other programmable data processor, or other device to perform a series of arithmetic steps on the computer, other programmable device, or other device. Generates a computer-implemented process such that an instruction executed on another programmable device, or other device, performs a specified function / action in one or more blocks of a flowchart or block diagram or both. Can be done.

The flowcharts and block diagrams in the figure show the architecture, functionality, and behavior of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagram may represent a module, segment, or part of an instruction that contains one or more executable instructions to perform a specified logical function (s). In some alternative implementations, the functions shown in the blocks can occur out of the order shown in the figure. For example, depending on the functionality involved, two blocks shown in succession can actually be executed at substantially the same time, or the blocks can optionally be executed in reverse order. Each block of the block diagram and / or flowchart diagram, and the combination of blocks in the block diagram and / or flowchart diagram, performs a specified function or action, or performs a combination of dedicated hardware and computer instructions. Also note that it can be implemented by a dedicated hardware-based system.

The computational components of FIG. 1, including the storage controller 100 and subsystem 112, may be implemented in one or more computer systems, such as the computer system 702 shown in FIG. The computer system / server 702 may be described in the general context of computer system executable instructions such as program modules executed by the computer system. In general, a program module can include routines, programs, objects, components, logic, data structures, etc. that perform a particular task or implement a particular abstract data type. The computer system / server 702 may be implemented in a decentralized cloud computing environment in which tasks are performed by remote processing devices linked over a communication network. In a distributed cloud computing environment, program modules can be located on both local and remote computer system storage media, including memory storage devices.

As shown in FIG. 7, the computer system / server 702 is shown in the form of a general purpose computing device. The components of the computer system / server 702 include one or more processors or processing units 704, system memory 706, and a bus 708 that connects various system components, including system memory 706, to processor 704. It may include, but is not limited to. Bus 708 is of several types of bus structures, including memory buses or memory controllers, peripheral buses, accelerated graphics ports, and processors or local buses that use any of the various bus architectures. Represents one or more of. By way of example, but not by limitation, such architectures include industry standard architecture (ISA) bus, Micro Channel Architecture (MCA) bus, extended ISA (EISA) bus, and video. -Includes the Video Electronics Standards Association (VESA) local bus and the Peripheral Component Interfaces (PCI) bus.

The computer system / server 702 typically includes various computer system readable media. Such media can be any available medium accessible by the computer system / server 702, including both volatile and non-volatile media, removable and non-removable media.

The system memory 706 can include computer system readable media in the form of volatile memory, such as random access memory (RAM) 710 and / or cache memory 712. The computer system / server 702 may further include other removable / non-removable, volatile / non-volatile computer system storage media. As just one example, a storage system 713 can be provided for reading and writing to a non-removable non-volatile magnetic medium (not shown, commonly referred to as a "hard drive"). Although not shown, removal of magnetic disk drives for reading and writing to and from removable non-volatile magnetic disks (eg, "flexible disks") and removal of CD-ROMs, DVD-ROMs, or other optical media, etc. An optical disk drive for reading and writing can be provided on a possible non-volatile optical disk. In such cases, each can be connected to bus 708 by one or more data media interfaces. As further illustrated and described below, memory 706 includes at least one program product having a set (eg, at least one) of program modules configured to perform the functions of embodiments of the present invention. obtain.

A program / utility 714 having a set (at least one) of program modules 716 is similar, but not limited to, an operating system, one or more application programs, other program modules, and program data. In addition, it can be stored in the memory 706. Each operating system, one or more application programs, other program modules, and program data or any combination thereof may include an implementation of a networking environment. The components of computer 702 can be implemented as program modules 716 that generally perform the functions and / or methodologies of the embodiments of the invention described herein. The system of FIG. 1 may be implemented in one or more computer systems 702, and when implemented in multiple computer systems 702, the computer systems may communicate over a network.

The computer system / server 702 also includes one or more external devices 718, such as a keyboard, pointing device, display 720, and one or more devices that allow the user to interact with the computer system / server 702. Alternatively, it may communicate with any device (eg, network card, modem, etc.) that allows the computer system / server 702 to communicate with one or more other computing devices, or a combination thereof. Such communication can be done via the input / output (I / O) interface 722. Furthermore, the computer system / server 702 is via a network adapter 724 to a local area network (LAN), a general wide area network (WAN), or a public network (eg, the Internet). Alternatively, it can communicate with one or more networks, such as a combination thereof. As shown, the network adapter 724 communicates with other components of the computer system / server 702 via bus 708. Although not shown, it should be understood that other hardware and / or software components may be used with the computer system / server 702. Examples include, but are not limited to, microcodes, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data archive storage systems.

"One embodiment", "Embodied", "Multiple embodiments", "The embodied", "The embodied", "The embodied", "The embodied" Unless otherwise specified, the terms "one or more embodied", "some embodiments", and "one embodied" are "the present invention". Means one or more (but not all) embodiments (s).

The terms "inclusion," "comprising," "having," and their variations, unless otherwise stated, mean "inclusion but not limited to." ..

The list of listed items does not mean that any or all of the items are mutually exclusive, unless otherwise stated.

The terms "a," "an," and "the" mean "one or more," unless otherwise stated.

Devices that communicate with each other do not need to communicate continuously with each other unless otherwise specified. In addition, devices communicating with each other may communicate directly or indirectly via one or more intermediaries.

The description of an embodiment having several components communicating with each other does not mean that all such components are required. Rather, a variety of optional components are described to illustrate a wide variety of possible embodiments of the present invention.

When a single device or article is described herein, it is easy for two or more devices / articles (with or without collaboration) to be used instead of a single device / article. It will be clear to. Similarly, if two or more devices or articles (with or without collaboration) are described herein, a single device / article may be used in place of the two or more devices or articles. , Or it will be readily apparent that different numbers of devices / articles may be used in place of the number of devices or programs shown. The functionality and / or features of the device may optionally be embodied by one or more other devices not explicitly described as having such functionality / features. As such, other embodiments of the invention need not include the device itself.

The aforementioned description of various embodiments of the present invention has been presented for purposes of illustration and description. It is neither exhaustive nor intended to be limited to the exact form in which the invention is disclosed. Many modifications and modifications are possible in the light of the above teachings. The scope of the present invention shall be limited not by this detailed description but by the claims attached thereto. The above specifications, examples, and data provide a complete description of the manufacture and use of the configurations of the present invention. The present invention is within the scope of the appended claims, as many embodiments of the invention can be made without departing from the scope of the invention.

Claims (25)

A computer program product for managing input / output requests to multiple storage devices, including a computer-readable storage medium that embodies computer-readable program code that can be executed to perform an operation. The operation is
To store a group of tracks stored in the storage device in the cache,
Granting an exclusive track lock to the track in the group in the cache for writing to a track in the group in the cache, wherein the exclusive track lock is in the cache. Can be held at the same time for writing to different tracks, granting and
Granting an exclusive group lock to the group of tracks in the cache in order to destage the track in the group from the cache to the storage device.
Releasing the exclusive group lock in response to completing the destage of the truck in the group in the cache to the storage device.
Computer program products, including. The above operation further
Receiving a write to the target track in the group
Determining if the exclusive group lock to the group is held and
In response to the determination that the exclusive group lock is not held, the exclusive track lock is given to the target track, and
Retrying access to the exclusive track lock for the target track in response to determining that the exclusive group lock is held.
The computer program product according to claim 1. The above operation further
Granting a shared group lock for writing to the track in the group in the cache.
Releasing the shared group lock for the write in response to the completion of writing to the track in the cache.
In response to initiating the destage of a group of said tracks, determining whether any shared group lock is held, said exclusive group lock is said to destage. Granted only in response to determining that there are no shared group locks held for the group.
The computer program product according to claim 1. The destage of the group of trucks further
Identifying modified data in tracks within said group in said cache and
Read from at least one of the storage devices by generating a read request indicating the modified data and data about the old parity data of the group and adding it to the submission queue to read from the cache. To staging and
Using the modified data staged in the cache and the read data for the old parity data to calculate new parity data.
Writing the new parity data and the modified data in the cache to the at least one of the storage devices, the exclusive group lock is the new parity data and the new parity data. Writing and releasing the modified data in response to the completion of the writing to at least one of the storage devices.
The computer program product according to claim 1. Writing the new parity data and the modified data
The at least of the storage devices from the cache by generating write requests indicating the modified data in the cache and data about the new parity data and adding them to the submission queue. Transfer to one,
4. The computer program product according to claim 4. The computer-readable program code is executed by a system that includes the computer-readable storage medium, the system is coupled to a controller that interfaces with the storage device, and the operation is further enhanced.
Setting the doorbell register of the submission queue when adding the read request and the write request to the submission queue, the setting of the doorbell register processes the read request. Then, data is transferred from the at least one of the storage devices to the cache, the write request is processed, and the changed data and the new parity data are transferred from the cache to the storage device. Signaling, setting, to forward to at least one of said controllers.
5. The computer program product according to claim 5. To generate the read request
Identifying modified segments of tracks within the group
Assigning a cache segment in the cache to the identified modified segment and the segment of old parity data,
To generate a list showing the segment of the modified data and the parity data and the destination position of the allocated cache segment in the cache, the list of which is the storage device. Producing, provided with the read request, for use in identifying the data to be read from said at least one of them and written to said destination location in said cache.
4. The computer program product according to claim 4. Destage the track further
Determining that all the tracks in the group are in the cache
Computing new parity data for the group from the track for the group in the cache.
Shows the modified track and the new parity data of the group in the queue and the destination position on at least one of the modified track and the storage device that writes the new parity data. To generate a write request containing a list and add it to the submission queue, the controller accesses the write request and the list in the submission queue and makes the changes shown in the list. Accessing the data and the new parity data of the track, writing and adding over at least one of the storage devices indicated as the destination location in the list.
The computer program product according to claim 1. The above operation further
Initiating the operation for staging track data into the cache from at least one of the storage devices.
Assigning a segment in the cache to the staging data and
Granting a lock to the assigned segment in the cache for the data to be staged,
Adding a read request to the submission queue indicating the data for the track to be staged in the cache.
To store the data from the read request in the allocated segment in the cache.
Releasing the lock on the segment in the cache in response to completing the read of the data for the segment in the cache.
The computer program product according to claim 1. A computer program product for managing input / output requests to tracks in cache stored in multiple storage devices, a computer-readable program that performs operations when executed by multiple processing cores. The operation includes a computer-readable storage medium in which the code is embodied.
Granting a shared stride lock to writes to a track in a stride group of tracks, said shared stride lock is granted to writes to write to different tracks in the stride group at the same time. To give and give
Determining if there is an unreleased shared stride lock for writing in response to a destage request to destage the stride group of said track.
Granting an exclusive stride lock for destage of a stride group of tracks from the cache to the storage device in response to determining that there are no unreleased shared stride locks for writing. When,
Computer program products, including. The above operation further
Communicating with the controller that manages access to the storage device through the fabric
In order to respond to a track read request, a read request from the storage device to stage the track into the cache is added to at least one submission queue, where the controller puts the fabric on. To retrieve the read request from the at least one submission queue via the storage device, read data from the storage device, write to the cache, and add.
Adding a write request to write the destaged track of the stride group to the at least one submission queue in response to receiving the exclusive stride lock to the stride group. The controller retrieves the write request from the at least one submission queue through the fabric and writes or adds data to the storage device.
10. The computer program product of claim 10. A system for managing input / output requests to multiple storage devices.
With at least one processor
The operation comprises a computer-readable storage medium embodying computer-readable program code that performs the operation when executed by the at least one processor.
To store a group of tracks stored in the storage device in the cache,
Granting an exclusive track lock to the track in the group in the cache for writing to a track in the group in the cache, wherein the exclusive track lock is in the cache. Can be held at the same time for writing to different tracks, granting and
Granting an exclusive group lock to the group of tracks in the cache in order to destage the track in the group from the cache to the storage device.
Releasing the exclusive group lock in response to completing the destage of the truck in the group in the cache to the storage device.
Including the system. The above operation further
Receiving a write to the target track in the group
Determining if the exclusive group lock to the group is held and
In response to the determination that the exclusive group lock is not held, the exclusive track lock is given to the target track, and
Retrying access to the exclusive track lock for the target track in response to determining that the exclusive group lock is held.
12. The system according to claim 12. The above operation further
Granting a shared group lock for writing to the track in the group in the cache.
Releasing the shared group lock for the write in response to the completion of writing to the track in the cache.
In response to initiating the destage of a group of said tracks, determining whether any shared group lock is held, said exclusive group lock is said to destage. Granted only in response to determining that there are no shared group locks held for the group.
12. The system according to claim 12. The destage of the group of trucks further
Identifying modified data in tracks within said group in said cache and
Read from at least one of the storage devices by generating a read request indicating the modified data and data about the old parity data of the group and adding it to the submission queue to read from the cache. To staging and
Using the modified data staged in the cache and the read data for the old parity data to calculate new parity data.
Writing the new parity data and the modified data in the cache to the at least one of the storage devices, the exclusive group lock is the new parity data and the new parity data. Writing and releasing the modified data in response to the completion of the writing to at least one of the storage devices.
12. The system according to claim 12. Writing the new parity data and the modified data
The at least of the storage devices from the cache by generating write requests indicating the modified data in the cache and data about the new parity data and adding them to the submission queue. Transfer to one,
15. The system of claim 15. To generate the read request
Identifying modified segments of tracks within the group
Assigning a cache segment in the cache to the identified modified segment and the segment of old parity data,
To generate a list showing the segment of the modified data and the parity data and the destination position of the allocated cache segment in the cache, the list of which is the storage device. Producing, provided with the read request, for use in identifying the data to be read from said at least one of them and written to said destination location in said cache.
15. The system of claim 15. A system for managing input / output requests to multiple storage devices.
With multiple processing cores
The operation comprises a computer-readable storage medium embodying computer-readable program code that performs the operation when executed by the processing core.
Granting a shared stride lock to writes to a track in a stride group of tracks, said shared stride lock is granted to writes to write to different tracks in the stride group at the same time. To give and give
Determining if there is an unreleased shared stride lock for writing in response to a destage request to destage the stride group of said track.
Granting an exclusive stride lock for destage of a track stride group from the cache to the storage device in response to determining that there is no unreleased shared stride lock for writing. ,
Including the system. The above operation further
Communicating with the controller that manages access to the storage device through the fabric
In order to respond to a track read request, a read request from the storage device to stage the track into the cache is added to at least one submission queue, where the controller puts the fabric on. To retrieve the read request from the at least one submission queue via the storage device, read data from the storage device, write to the cache, and add.
Adding a write request to write the destaged track of the stride group to the at least one submission queue in response to receiving the exclusive stride lock to the stride group. The controller retrieves the write request from the at least one submission queue through the fabric and writes or adds data to the storage device.
18. The system of claim 18. A method for managing input / output requests to multiple storage devices.
To store a group of tracks stored in the storage device in the cache,
Granting an exclusive track lock to the track in the group in the cache for writing to a track in the group in the cache, wherein the exclusive track lock is in the cache. Can be held at the same time for writing to different tracks, granting and
Granting an exclusive group lock to the group of tracks in the cache in order to destage the track in the group from the cache to the storage device.
Releasing the exclusive group lock in response to completing the destage of the truck in the group in the cache to the storage device.
Including methods. Receiving a write to the target track in the group
Determining if the exclusive group lock to the group is held and
In response to the determination that the exclusive group lock is not held, the exclusive track lock is given to the target track, and
Retrying access to the exclusive track lock for the target track in response to determining that the exclusive group lock is held.
20. The method of claim 20. Granting a shared group lock for writing to the track in the group in the cache.
Releasing the shared group lock for the write in response to the completion of writing to the track in the cache.
In response to initiating the destage of a group of said tracks, determining whether any shared group lock is held, said exclusive group lock is said to destage. Granted only in response to determining that there are no shared group locks held for the group.
20. The method of claim 20. The destage of the group of trucks further
Identifying modified data in tracks within said group in said cache and
Read from at least one of the storage devices by generating a read request indicating the modified data and data about the old parity data of the group and adding it to the submission queue to read from the cache. To staging and
Using the modified data staged in the cache and the read data for the old parity data to calculate new parity data.
Writing the new parity data and the modified data in the cache to the at least one of the storage devices, the exclusive group lock is the new parity data and the new parity data. Writing and releasing the modified data in response to the completion of the writing to at least one of the storage devices.
20. The method of claim 20. Writing the new parity data and the modified data
The at least of the storage devices from the cache by generating write requests indicating the modified data in the cache and data about the new parity data and adding them to the submission queue. Transfer to one,
23. The method of claim 23. To generate the read request
Identifying modified segments of tracks within the group
Assigning a cache segment in the cache to the identified modified segment and the segment of old parity data,
To generate a list showing the segment of the modified data and the parity data and the destination position of the allocated cache segment in the cache, the list of which is the storage device. Producing, provided with the read request, for use in identifying the data to be read from said at least one of them and written to said destination location in said cache.
23. The method of claim 23.

Images